Project Title: Direct Numerical Simulations of Turbulent Flows in Cylindrical Coordinate Systems
Project Chief Investigator: Dr Jun-De Li
Funding Obtained: $20,000
Project Summary: Direct numerical simulation (DNS) of turbulent flows started in the late 1970's and became popular in the 1980's and 1990's. In DNS, the governing equations for turbulence (Navier-Stokes equations) are solved numerically down to the respective smallest length &d time scales without using turbulence models. Because of the large number of length and time scales involved, it requires very high speed computers with large memories, and it has only been applied to 3D turbulent flow fields with periodical. boundary conditions in at least two directions such as isotropic turbulence and turbulent channel flows. Also because of the high accuracy and fast convergence required, the commonly used finite difference and finite element numerical schemes are generally not adequate, and spectral method is the preferred choice in DNS. In performing DNS, the main limitation in speed is solving Poisson type equations. In this project, the fast Poisson solver in cylindrical coordinate systems with spectral accuracy developed by the Chief Investigator recently will first be coded using C (or Fortran) language.

Various boundary conditions will be incorporated into the code so that it can be developed as a generic solver to be used not only for DNS but also for applications in other fields. Expertises at VPAC will be required to parallelise the code so that it can be run in cluster of computers at high speed. The code will then be used as a base for developing DNS code for simulating turbulent flows in cylindrical coordinate systems such as flows in pipes and between rotating cylinders (Taylor-Couette flows). The codes will be run at different Reynolds numbers to study: (1) the transition of laminar I flows to turbulent flows in the two different flow geometries, (2) the Reynolds number effects in low Reynolds number turbulent flows, and (3) the highest Reynolds numbers can be simulated for turbulent flows in pipes and rotating cylinders using the computation capability of VPAC and APAC.

Project Title: Modelling coral atoll hydrodynamics and habitat suitability to inform sustainable human resettlement of Rongelap Atoll, Republic of the Marshall Islands
Project Chief Investigator: Eric Peterson
Funding Obtained: $20,067
Project Summary: Working with coral reef specialists in the Marshall Islands based non-government organization NRAS-conservation, I propose to use VPAC facilities and the Commonwealth Bureau of Meteorology and CSIRO "BLUElink> Ocean Forecasting" research products to map and simulate the atmospheric and hydrodynamic microclimates of Rongelap atoll to classify habitat suitability zones. This work employs previous findings of the interactions between physical oceanography and biodiversity, and the outcome will be to inform marine protected area design by the Rongelap atoll local government RalGov. The model will be validated by natural resource assessment surveys conducted in the Republic of the Marshall Islands 2002-2006 and findings from Woods Hole Oceanographic Institute.

Application Hydrodynamics may explain what drives larval distribution in and around an isolated atoll and what are the physical qualities that are important in ecological function and habitat suitability. This project aims to complement the concurrent long-term monitoring of coral reef resources and biodiversity at Rongelap Atoll, to inform establishment of a system of Marine Reserves at strategic locations around the Atoll. Rongelap Atoll is a far northern member of the western Ralik Island Chain in the Marshall Islands, and part of Micronesia, listed by the Australian Government as an adjacent developing country and a global biodiversity hotspot. Within the Micronesian hotspot, Rongelap Atoll is unique because it has been virtually uninhabited for 5 decades. The Atoll's isolated reefs have not been subjected to subsistence collecting or fishing for 50 years, because the local inhabitants were evacuated after radioactive fallout from atomic tests at Bikini Atoll in 1954. In 1998, the Rongelap Atoll Local Government (RalGov) initiated a resettlement program that involved decontamination works and the building of infrastructure. With the impending resettlement of Rongelapese people, this project will present a unique opportunity to work cooperatively with RalGov and other local community groups to map habitat suitability (according to hydrodynamics) and manage the biodiversity of the Atoll's coral reefs in a proactive and sustainable way.

Project Title: Web-based hospital management using discrete event simulation and data visualization
Project Chief Investigator: Gitesh Raikundalia
Funding Obtained: $12,268
Project Summary: Hospital simulation is an area that needs to be addressed if hospitals are to be managed efficiently and effectively. Hospital management decision-making needs to deal with several major factors, such as patient flow, staffing and clinical resources. One of the investigators has developed a tool, G2 Exchange, that simulates scenarios at Royal Melbourne Hospital, and the tool has received positive feedback from stakeholders. A Web version of G2 Exchange, e-Simulator, has also been developed by the investigators. Although e-Simulator employs the familiar Web page model, it generates enormous quantities of data that need hours to be processed for relevant simulations to be generated for use by hospital managers. Therefore, advanced computing is vital and fundamental in generating hospital simulations in a matter of minutes in the time-critical environment of a hospital. Such a powerful system has not been developed by related research. The system to be developed by the investigators is unique in that it will apply to multiple hospitals and be relevant to non-technical staff as well as hospital managers. This project aims to develop techniques (algorithms) for generating hospital simulations from very large data sets and implementing these algorithms within software. The software will be integrated with e-Simulator, so that a hospital manager will use the Web-based e-Simulator and view appropriate simulations for use in decision- making.

The project also aims to evaluate the effectiveness of the final system with hospital managers. This will provide lessons learnt in developing such a system and allow us to further improve the final system for real-world use in hospitals. The expected outcomes of this project include the following: A simulation system that can assist hospital managers in very quick time to carry out hospital decision- making. Such a system is very important considering the vital role that hospitals play in the community, yet it can be developed in a number of months at a low cost. The quality of this project is supported by its basis on G2 Exchange, which has been shown to be effective in Royal Melbourne Hospital and has also been published. Given both the tremendous importance and ubiquity of hospitals in society, the publication of the system to inform hospital managers within and outside Australia is essential. High-quality publications must be published in to reach as many hospital managers as possible. Generation of further projects that extend this work. It is possible to carn out case studies on the software in particular hospitals to learn more about the use and effectiveness of the software. Since solution of this relevant problem requires advanced computing, without which hospital managers cannot make important decisions in a short amount of time, it is clear that VPAC's computing facilities will be used effectively and applied to a highly essential area (health, or specifically, e-Health).

Project Title: In search of grid independent results for the simulation of an ISO 9705 room fire using the CFD model, FDS

Project Chief Investigator: Professor. Ian Thomas

Funding Obtained: $9,332
Project Summary: The aim of this e-research project is to obtain grid independent results when conducting numerical simulation of an ISO 9705 room fire experiment using a computational fluid dynamics (CFD) model commonly used by fire safety engineers in Australia and overseas. To evaluate the performance of fire safety designs, fire and smoke modelling is a highly desirable tool for engineers due to the high cost of experimental performance studies. In the current era of fast computers, the use of CFD is possible for real life fire scenarios. The Fire Dynamics Simulator (FDS) developed at the National Institute of Standards and Technology (NIST), US, is such a CFD model. This model has been widely used by fire engineers due to some of its unique features (the ability to predict the course of a fire through ignition, growth, established burning and decline through to extinguishment or burnout) and its availability at no cost from NIST. The FDS model incorporates a Large Eddy Simulation (LES) turbulence model, mixture-fraction combustion model, and finite-volume radiative heat transfer model. In conducting CFD analysis it is essential to undertake a process of grid refinement in which the grid spacing (cell size) used in the analysis is gradually reduced and the effect of the reduced grid spacing on the predicted outcome is examined. It is usual to find that as the grid spacing is reduced the results converge. Thus further reducing the grid spacing has virtually no effect on the results produced, inferring the achievement of grid convergence.

To obtain any meaningful result, it is imperative to achieve grid independence while simulating a fire scenario, not just using arbitrary cell size. Such numerical testing was attempted for a liquid fuel fire within an ISO 9705 room and it was found that, even after using cell sizes down to 12.5 mm, convergence had not been achieved. When an attempt was made to reduce the cell size below 12.5 mm, the PC could not run the simulation. When contacted, the FDS developer acknowledged that obtaining grid convergence is a major challenge for FDS. The developers postulated that by modifying the stoichiometric ratio it is possible to get a reasonable approximation to the heat release rate (HRR) of the fire, even when the fire is not well-resolved. Such an automatic modification routine is included in FDS. However, this routine may fail if the grid resolution is too coarse. Most fire safety engineers are not aware of this limitation. While simulating a fire scenario using the FDS model, most fire safety engineers use computational cell size that can only be supported by their computing resources. The objective of this project is to evaluate how numerical results for the simulation of a simple fuel package (liquid) fire in an ISO 9705 room vary with the change of computational cell size and what computer resources are required to obtain a grid-independent result for this fire scenario. From these numerical data an analysis of the applicability of results to all compartment fires will also be carried out with an intention to propose a universal non-dimensional profile. This proposed profile will be extremely useful for fire safety engineers in making a judgment of their fire simulation results obtained using computationally viable grids.

Project Title: Semi-automating the process of marking up sentencing decisions in order to use knowledge discovery to analyse the sentencing process.

Project Chief Investigator: Professor John Zeleznikow

Funding Obtained: $13,333
Project Summary: It is proposed to develop a plea-bargaining system to support sentencing negotiations between Victoria Legal Aid (VLA) defence counsel and the Victorian Office of Public Prosecution. (Stranieri and Zeleznikow 2005) detail on-going research in the domain of knowledge discovery from legal databases. A 2000 VPAC grant, the automated detection of change in concepts by analysing data, investigated how (VLA) distributes its legal aid grants. The research indicated to VLA how it could appropriately use its funds. It was the basis of the GetAid system which is being used by VLA to determine eligibility for legal aid (see (Stranieri et a1 2000), (Stranieri and Zeleznikow 2000) and (Ivkovic et a1 2003).

To construct our plea-bargaining system we are utilising a modified Touhin argument structure to construct the relevant ontology. The sentencing ontology is being elicited from domain experts. There are approximately 70 factors that a judge may take in to account when making a decision. Given our large data set (the 2004-5 year consisted of 438 judgments, each being approximately 10 pages), we wish to use a data mining technique, such as association rules, to learn relationships between the variables involved in the determination of a sentence.

Concurrent with this exploration we would like to investigate the possibility of semi-automating the process of marking up the sentencing decisions. Using an open source software package TAMS analyser (http://tarnsvs.sourceforge.net/), the written sentencing decisions can be marked up. For each sentencing decision, an attempt is made to identify as many factors as possible. This is currently a manual process. It would be extremely useful to m o w this software to enable the semi-automation of this task. It is necessary to retain the generic concept of this program so that it can be made part of a tool box for building sentencing decision support systems in other jurisdictions.

Support from VPAC will allow us to:

I 1. Develop the TAMS tool to enable the semi-automation of the collection of the sentencing factors from written decisions.

2.2. Support the analysis of the large free text sentencing data set by association rules and possibly other valid data mining techniques.